Face recognition based on feature vector generation using Centric mean with forward and backward diagonal mean using DCT

In the modern era the world comes nearer to every individual as an IT revolution where all the applications are computerized. As the level of security breaches and frauds of transaction increases, it requires highly secure identification and personal verification technologies. Here user identity plays a very important role to uniquely verify or authenticate the individual person. Instead of designing more complex system, which is more expensive and which requires more software and hardware resources, it is essential to think about to bridge the gap which will create a system with simplicity, less costly and efficient, as well as socially acceptable. The factors which highly impact the face recognition system performance are illumination and pose variations. Feature extraction is the key to reach face recognition. In literature various feature extraction techniques are available. The paper presents six different techniques to generate a feature vector by calculating Centric mean and forward and backward diagonal mean and then apply DCT over it which will reduce the feature vector size and yields less computational complexity giving faster recognition. Experimental tests performed on the ORL and UMIST database achieves a recognition rate of 99 percent, with low computational cost without any pre-processing step.